CN117633640A - Unmanned aerial vehicle track prediction method based on flight mode identification - Google Patents
Unmanned aerial vehicle track prediction method based on flight mode identification Download PDFInfo
- Publication number
- CN117633640A CN117633640A CN202311576511.3A CN202311576511A CN117633640A CN 117633640 A CN117633640 A CN 117633640A CN 202311576511 A CN202311576511 A CN 202311576511A CN 117633640 A CN117633640 A CN 117633640A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- data
- flight
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000013528 artificial neural network Methods 0.000 claims abstract description 25
- 230000033001 locomotion Effects 0.000 claims abstract description 19
- 238000012706 support-vector machine Methods 0.000 claims abstract description 10
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 24
- 210000002569 neuron Anatomy 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000007781 pre-processing Methods 0.000 claims description 15
- 230000002159 abnormal effect Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 14
- 238000006073 displacement reaction Methods 0.000 claims description 12
- 230000005540 biological transmission Effects 0.000 claims description 11
- 238000011478 gradient descent method Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000003909 pattern recognition Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 8
- 230000009194 climbing Effects 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 6
- 238000004891 communication Methods 0.000 claims description 6
- 238000007405 data analysis Methods 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 230000008030 elimination Effects 0.000 claims description 5
- 238000003379 elimination reaction Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000012546 transfer Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 230000006872 improvement Effects 0.000 abstract description 4
- 238000011161 development Methods 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 10
- 238000010801 machine learning Methods 0.000 description 6
- 238000011160 research Methods 0.000 description 6
- 239000002245 particle Substances 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- RZVHIXYEVGDQDX-UHFFFAOYSA-N 9,10-anthraquinone Chemical compound C1=CC=C2C(=O)C3=CC=CC=C3C(=O)C2=C1 RZVHIXYEVGDQDX-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013079 data visualisation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000000575 pesticide Substances 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method for predicting a flight path of an unmanned aerial vehicle by using flight mode identification, which is characterized in that a flight path prediction system of the unmanned aerial vehicle is established aiming at different flight states, the motion characteristics of the unmanned aerial vehicle are established through a principal component analysis method, the flight states of the unmanned aerial vehicle are identified by using an improved support vector machine model, and the flight paths of the unmanned aerial vehicle are predicted by respectively establishing prediction models according to the different flight states. The method adopts a support vector machine model based on binary tree improvement to identify different flight states of the unmanned aerial vehicle; and establishing a track prediction model of different flight states by adopting a BP neural network. Not only the historical flight path of the unmanned aerial vehicle is considered, but also the current flight state of the unmanned aerial vehicle is considered, the engineering problem of insufficient prediction accuracy caused by the lack of consideration of the current flight state of the unmanned aerial vehicle in unmanned aerial vehicle flight path prediction is solved, a more accurate prediction method in the unmanned aerial vehicle flight path prediction field is provided, and the positioning and navigation technology development of the autonomous aircraft in the future aviation field is facilitated.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicles, in particular to an unmanned aerial vehicle track prediction method.
Background
The unmanned aerial vehicle has the advantages of convenient deployment, strong maneuverability, low cost and the like. With the rapid development of the unmanned aerial vehicle field, the unmanned aerial vehicle has great value in the military, civil and commercial fields. In the military field, unmanned aerial vehicle can be used to accomplish tasks such as reconnaissance, target location, signal information search, prevent and control firepower bait. In the civil field, unmanned aerial vehicle can accomplish high-altitude high-risk work such as express delivery transportation, pesticide spray, antenna inspection. In the commercial field, unmanned aerial vehicle can accomplish multiple angle aerial photography, shoots the photographic work that has more viewing value.
With the rapid rise of the unmanned aerial vehicle field, the safety problem of the unmanned aerial vehicle is increasingly highlighted, and an unmanned aerial vehicle safety supervision system is generated. The unmanned aerial vehicle track prediction refers to a process of predicting future tracks of the unmanned aerial vehicle through unmanned aerial vehicle local information, and more accurate navigation data can be provided for unmanned aerial vehicle control and navigation. The unmanned aerial vehicle track prediction is a core technology of unmanned aerial vehicle safety supervision, is a precondition for unmanned aerial vehicle control and navigation, and is also a necessary condition for unmanned aerial vehicle autonomous flight.
In the field of unmanned aerial vehicle track prediction, the unmanned aerial vehicle track prediction method mainly comprises the following three steps: hybrid estimation, particle motion model, and machine learning. The hybrid estimation algorithm is firstly suitable for the field of positioning tracking, and mainly comprises Kalman filtering, particle filtering and the like. And the unmanned aerial vehicle particle motion model regards the unmanned aerial vehicle as moving particles in the motion process, ignores the rolling motion of the unmanned aerial vehicle, and builds an unmanned aerial vehicle track module by using a full energy equation. And predicting the flight path of the current unmanned aerial vehicle by learning the airborne data and the flight path change in the historical navigation process of the unmanned aerial vehicle based on a machine learning method. However, most of the researches on unmanned aerial vehicle track prediction are carried out on unmanned aerial vehicle historical track researches, consideration of the unmanned aerial vehicle real-time flight state is insufficient, and a prediction method for predicting the unmanned aerial vehicle track aiming at the unmanned aerial vehicle real-time flight state is not yet found.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a method for predicting the flight path of an unmanned aerial vehicle by identifying a flight mode. In order to solve the problem of insufficient consideration of the flight state of the unmanned aerial vehicle in the unmanned aerial vehicle flight path prediction process, the invention discloses a method for performing unmanned aerial vehicle flight path prediction by flight mode identification, and solves the problem of unfocused predicted targets in the current unmanned aerial vehicle flight path prediction process. The flow chart of the unmanned aerial vehicle track prediction method disclosed by the invention is shown in fig. 1, and comprises the following steps: analyzing basic parameters collected in the flight process of the unmanned aerial vehicle; preprocessing data of acquisition parameters of the unmanned aerial vehicle; collecting the accumulated contribution rate of the motion parameters and calculating; constructing motion characteristics in the flight process of the unmanned aerial vehicle; classifying the flight state of the unmanned aerial vehicle by using an improved support vector machine model; and constructing a corresponding unmanned aerial vehicle track prediction model aiming at each flight state, finally obtaining a short-term predicted track of the unmanned aerial vehicle, and finishing track prediction of different flight states of the unmanned aerial vehicle. The method can effectively identify the flight state of the unmanned aerial vehicle and forecast the flight path.
The technical scheme adopted by the invention for solving the technical problems comprises the following steps:
a method for predicting unmanned aerial vehicle flight path by flight pattern recognition comprises the following steps:
step 1: collecting and analyzing airborne data of the unmanned aerial vehicle;
step 2: preprocessing unmanned aerial vehicle track data;
because noise exists in the unmanned aerial vehicle track data, data preprocessing is needed before unmanned aerial vehicle data analysis, abnormal data is removed in the data preprocessing, interpolation processing is carried out on the missing data, and data is output after the data preprocessing flow is completed;
step 3: constructing the motion characteristics of the unmanned aerial vehicle;
in the machine learning field, excessive learning directly affects classification accuracy, and therefore, it is necessary to perform data dimension reduction. The principal component analysis method (Principal Component Analysis, PCA) can be used as a dimension reduction algorithm to reduce a plurality of indexes into a plurality of principal components, the principal components are formed by linearly combining original variables and have no relation, and most of useful information in the original data can be embodied;
step 4: unmanned aerial vehicle flight mode identification;
data classification by finding hyperplane that segments data for a given one of the classified data setsWherein y is i E { +1, -1}, if the two samples are linearly separable, there is one hyperplane:
ω T x+b＝0
the two types of samples can be separated, then there is y for each sample i (x T x+b) > 0, the distance of the sample to the segmentation hyperplane is:
wherein ω is a hyperplane parameter;
defining γ as the shortest distance from all samples in the dataset to the segmentation hyperplane, then:
searching a most suitable hyperplane to classify data, and when the hyperplane meets the maximum interval gamma of all samples, dividing two data sets of the hyperplane segmentation is most stable and has strong robustness, namely solving an optimization equation:
wherein the method comprises the steps ofSatisfies the terms of ω x γ=1, i=1, 2, n., optimization equation reduction is:
the support vector machine model can realize classification recognition; the flight states of the unmanned aerial vehicle comprise five flight states of flat flight, turning, climbing, descending and spiraling, and the flight data of the unmanned aerial vehicle covering the five flight states are subjected to four-time two-classification, namely, the flight states of the unmanned aerial vehicle are subjected to multi-classification identification based on a support vector machine model improved by a binary tree, so that the five flight states of the unmanned aerial vehicle, namely, the climbing, flat flight, turning, spiraling and descending of the unmanned aerial vehicle can be accurately identified, and the unmanned aerial vehicle flight mode identification is realized;
step 5: predicting the flight path of the unmanned aerial vehicle;
the topology diagram of the BP neural network of the unmanned aerial vehicle track prediction model established by the invention is shown in figure 6, the unmanned aerial vehicle track prediction neural network selects 7 input layer variables and 4 output layer variables, and the 7 input layer variables are the positions (longitudinal displacement, transverse displacement and vertical displacement) in the directions of x, y and z coordinate axes, the angular motion parameters (pitch angle, roll angle and yaw angle) and the course respectively; the 4 output layer variables are displacement changes in the x, y and z axis directions and time, and the four output variables are adopted to describe a 4D track of the unmanned aerial vehicle;
and calculating the flight information of the unmanned aerial vehicle by the trained learning network to obtain a prediction result of the unmanned aerial vehicle flight path, thereby realizing unmanned aerial vehicle flight path prediction of flight pattern recognition.
The invention also provides an unmanned aerial vehicle-mounted data acquisition system, the system structure block diagram is shown in figure 2, the system comprises a lower computer, a wireless transmission module and an upper computer, the lower computer comprises a power supply module, a power module, a GPS positioning module, a flying height measuring module, an attitude detecting module and an STM32F427 singlechip main control module, the wireless transmission module is built by adopting a 3DR wireless data transmission module, and a wireless communication channel is provided for the wireless communication between the upper computer and the lower computer; the upper computer builds a monitoring system by LabVIEW, and realizes the functions of monitoring, controlling and recording airborne data in the flight process of the unmanned aerial vehicle.
In the unmanned aerial vehicle airborne data acquisition system, a power supply module provides power for a STM32F427 singlechip main control module and is matched with a power module to provide power for normal flight of the unmanned aerial vehicle, and a GPS positioning module measures the flight longitude and latitude of the unmanned aerial vehicle; the flying height measuring module is used for measuring the flying altitude of the unmanned aerial vehicle; the gesture detection module is used for measuring the flight acceleration and the angular speed of the unmanned aerial vehicle; the flight longitude and latitude, the flight altitude, the flight acceleration and the angular velocity form real-time flight information of the unmanned aerial vehicle, the real-time flight information is transmitted to an STM32F427 singlechip main control module, the STM32F427 singlechip main control module realizes wireless acquisition of unmanned aerial vehicle data through a wireless transmission module, the data is transmitted to a LabVIEW upper computer through a serial port, and the upper computer displays the flight altitude, the flight speed, the flight distance, the rotating speed of a propeller, the electric quantity of a battery and the flight attitude of the unmanned aerial vehicle in real time and records and stores real-time positions, flight tracks and historical data; and finally, inputting unmanned aerial vehicle-mounted information recorded and stored by the acquisition system into MATLAB software for subsequent data analysis.
In the step 2, the specific steps are as follows:
(1) removing abnormal data;
data n samples { x } are returned for a group of unmanned aerial vehicles 1 ，x 2 ，...，x n After uploading the returned data, calculating a first order value X i The method comprises the following steps:
X i ＝x i+1 -x i i＝1，2，...，n-1
X i describes the change of corresponding data of unmanned plane in a period of timeBy counting the cases X i The statistical characteristics of the data are removed, and the removing steps are as follows:
unmanned aerial vehicle flight sample x i Is a level difference value data X of (1) i The mathematical expectation of (a) is:
first order difference value data X i The standard deviation sigma of (2) is:
according to the statistical principle, the first order value X of the data sample i Obeying a normal distribution, according to the 3 sigma criterion, for X i Data lying within the (EX-3 sigma, EX+3 sigma) interval is considered normal data and is retained for X lying outside the (EX-3 sigma, EX+3 sigma) interval i Considered as abnormal data and culled, i.e. X outside the culling interval i Anomaly data and corresponding x i+1 Thereby realizing the elimination of abnormal data;
(2) missing data interpolation
The method comprises the steps of arranging and combining data after abnormal data are removed, carrying out interpolation smoothing processing on missing data by using a Lagrange interpolation method, and finally outputting repaired data, wherein the method comprises the following specific steps of:
the data arrangement after the outlier removal is expressed as (x) 0 ，y 0 )，(x 1 ，y 1 )，...，(x j ，y j )，...，(x k ，y k ) Wherein x is j Representing the position of the argument, y j The value of the function at this position is represented, j=0, 1,2,.. j All are different, and according to the Lagrange interpolation formula, the Lagrange interpolation polynomial is obtained as follows:
wherein each l j (x) For the Lagrangian base polynomial, the expression is:
lagrangian base polynomial l j (x) Is characterized in that x is j Take up a value of 1 for x of i.noteq.j i The upper value is 0;
and using the fitted Lagrangian polynomial as the estimation of the data with the outlier removed, and inserting the data position of the missing value into the Lagrangian polynomial to predict the value of the function at the position, thereby realizing the data interpolation estimation of the missing data position.
In the step 3, performing data dimension reduction on the data acquired by the unmanned aerial vehicle by adopting a principal component analysis method;
p indexes and n groups of samples of the preprocessed unmanned aerial vehicle flight data form an n multiplied by p matrix x as follows:
calculating the mean value of matrix x by columnThe method comprises the following steps:
calculated standard deviation S j The method comprises the following steps:
data is standardized, and standardized data are:
the normalized matrix for the original sample is obtained as:
the covariance matrix is calculated as:
wherein r is ij The method comprises the following steps:
solving the eigenvalue and eigenvector of covariance matrix R as lambda respectively i And a i I=1, 2, p. p is the index number of the sampling data, the value range of k is [1, p]；
The contribution rate c of the ith component to the total components i The definition is as follows:
cumulative contribution rate C i The definition is as follows:
the contribution rate c is firstly calculated i Sequencing from big to small, and sequentially accumulating the contribution rates to calculate an accumulated contribution rate C i Will accumulate the contribution rate C i The first more than 90% of the components are included as the main components.
In the step 4, the propagation process of the BP neural network algorithm comprises two aspects, namely forward propagation of the input layer information and reverse propagation of the output layer error in sequence;
(1) forward propagation of input layer information
As shown in fig. 6, the BP neural network is composed of an input layer, an hidden layer, and an output layer. The input layer has 7 neurons, the output layer has 4 neurons, and the hidden layer takes 4-14 neurons; the forward propagation of the information of the input layer is realized by transmitting the data weighting operation of the input layer to the hidden layer and then transmitting the data weighting operation of the hidden layer to the output layer, thereby realizing the forward propagation of the information of the neural network;
let the connection weight between the input layer (I) th neuron and the hidden layer (h) th neuron be v Ih The connection weight between the h neuron of the hidden layer and the J neuron of the output layer is omega hJ Then the input received by the h neuron of the hidden layer isThe input received by the J-th neuron of the output layer is +.>
Let the transfer function from the input layer to the hidden layer be f 1 Implicit layer to output layer transfer function of f 2 . Thus, the output of the hidden layer node is f 1 α h The output result of the output layer node is f 2 β J ；
(2) Counter propagation of output layer errors
Neural network parameter learning using a random gradient descent method for a given i-th set of samples (x i ，y i ) Inputting the sample into a neural network model, and obtaining network output as follows:
(a) Loss function
The loss function is used to quantify the difference between model predictions and real labels, defining a square type loss function as:
the total error generated for the p sets of neural network samples is defined as:
(b) Empirical risk minimization criteria
The smaller the model predicted loss, the stronger the model's predictive ability, in order to obtain a smaller expected error, the Empirical Risk (Empirical Risk) is calculated, i.e. the average loss of the training set is:
(c) Random gradient descent method
And optimizing an experience risk function by using a random gradient descent method, and continuously iterating the algorithm by training the random gradient descent method until the set precision is met, wherein the random gradient descent is converged to a local optimal solution after the iteration is ended.
The training steps of the random gradient descent method are as follows:
step 1: input training setVerification set V and learning rate α;
step 2: carrying out random initialization of a sample set;
step 3: calculation ofObtaining a gradient;
step 4: calculation ofThe result is taken asUpdating the parameter θ for the initial value of θ, i.e. +.>
Step 5: repeatedly performing step 3 and step 4 continuously until the model f (x i The method comprises the steps of carrying out a first treatment on the surface of the θ) when the error rate of the verification set V is the lowest, the iteration ends, and the loop is jumped out;
step 6: and outputting theta.
The invention has the beneficial effects that the current research situation at home and abroad shows that the improvement of the unmanned aerial vehicle track prediction method is mostly focused on the research basis of the historical track, and the research of the classification track prediction developed from the research angle of the real-time flight state of the unmanned aerial vehicle is not seen yet.
The invention predicts classified flight paths for five flight states (flat flight, turning, climbing, descending and hovering) of the unmanned aerial vehicle. Identifying different flight states of the unmanned aerial vehicle by adopting a support vector machine model based on binary tree improvement; and establishing a track prediction model of different flight states by adopting a BP neural network. According to the method, not only is the historical flight path of the unmanned aerial vehicle considered, but also the current flight state of the unmanned aerial vehicle is considered, the engineering problem of insufficient prediction accuracy caused by the lack of consideration of the current flight state of the unmanned aerial vehicle in unmanned aerial vehicle flight path prediction is solved, a more accurate prediction method in the unmanned aerial vehicle flight path prediction field is provided, an effective technical method can be provided for unmanned aerial vehicle autonomous flight control, and the positioning and navigation technical development of autonomous aircrafts in the future aviation field are facilitated.
Drawings
FIG. 1 is a flow chart of the operation of the present invention.
Fig. 2 is a block diagram of an information acquisition system of the unmanned aerial vehicle of the present invention.
Fig. 3 (a) is a visual image of the original data, and fig. 3 (b) is a visual image of the preprocessed data.
FIG. 4 is a statistical chart of the principal component contribution rate of the present invention.
Fig. 5 is a support vector machine classification diagram of the present invention.
Fig. 6 is a topology of a BP neural network of the present invention.
FIG. 7 is a diagram of a flat flight path prediction in accordance with the present invention.
Fig. 8 is a curve state track prediction diagram of the present invention.
FIG. 9 is a predicted view of the climb status track of the present invention.
FIG. 10 is a view of the descent state track prediction of the present invention.
FIG. 11 is a predicted view of a hover state flight path according to the present invention.
Fig. 12 is a graph of the track prediction error for different motion states according to the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
The invention aims to provide a method for predicting a flight path of an unmanned aerial vehicle by using flight mode identification.
Aiming at the existing problems, the invention mainly works sequentially with the following four working targets:
1) Preprocessing unmanned aerial vehicle flight data;
2) Constructing unmanned aerial vehicle motion characteristics;
3) Identifying the flight state of the unmanned aerial vehicle;
4) And carrying out classified flight path prediction aiming at different flight states of the unmanned aerial vehicle.
The work implementation steps of the above four work targets include the following five steps:
step 1: unmanned aerial vehicle flight data acquisition
The unmanned aerial vehicle flight data acquisition system is shown in the structural block diagram of fig. 2, and consists of a lower computer, a transmission system and an upper computer. The lower computer consists of a power supply module, a power module, a GPS positioning module, a flying height measuring module, a gesture detecting module and a main control module; the transmission module is built by adopting a 3DR wireless data transmission module, and provides a wireless communication channel for wireless communication between the upper computer and the lower computer; the upper computer builds a monitoring system by LabVIEW, so that the monitoring function of the unmanned aerial vehicle in the flight process is realized.
The flight altitude measurement system adopts an LPS331 barometer module to measure the atmospheric pressure of the current flight environment of the unmanned aerial vehicle. The barometric pressure-altitude formula is:
wherein P is 0 Is the standard atmospheric pressure of the sea level and is 101.325kPa; h is altitude; l is the temperature ramp down rate, which is about 0.0065K/m in dry air; t (T) 0 Is sea level standard temperature; g is the earth surface gravitational acceleration, about 9.8m/s2; m is the molar mass, about 0.0289644kg/mol; r is a universal gas constant of about 8.31447J/mol.K.
The available altitude calculation formula is:
taking the altitude of the unmanned aerial vehicle before taking off as the initial altitude h of the unmanned aerial vehicle 0 The difference between the altitude measured in the flight process of the unmanned aerial vehicle and the initial altitude is the flight altitude of the unmanned aerial vehicle. Namely:
H(t)＝h(t)-h 0
wherein H (t) is the altitude measured in the flight process of the unmanned aerial vehicle, H0 is the altitude measured before the unmanned aerial vehicle takes off, and H (t) is the real-time flight height measured by the unmanned aerial vehicle.
The unmanned aerial vehicle gesture detection module detects the spatial gesture data of the unmanned aerial vehicle through the MPU6050 gyroscope, and calculates the Euler angle formed by the unmanned aerial vehicle in the three-dimensional space and the coordinate axis by using the quaternion method. The Euler angle calculation flow is as follows:
firstly, initializing a gyroscope and calibrating an initial space coordinate system of the gyroscope; secondly, collecting data of a gyroscope device, and solving a quaternion differential equation according to the data; thirdly, collecting accelerometer data, and performing complementary filtering according to the accelerometer and gyroscope data to reduce errors; and finally, calculating the Euler angle of the current gesture of the unmanned aerial vehicle.
And inputting the acquired unmanned aerial vehicle-mounted information into MATLAB software for subsequent data analysis.
Step 2: preprocessing of unmanned aerial vehicle flight data
Because of the noise in the unmanned aerial vehicle flight data, it is necessary to perform data preprocessing prior to unmanned aerial vehicle flight data analysis. The data preprocessing mainly comprises the steps of removing abnormal data and interpolating missing data, and outputting the data after interpolation is completed.
Taking a group of unmanned aerial vehicle flight data acquisition systems to acquire and transmit back the real-time data of x, y and z axes of an upper computer as an example, carrying out unmanned aerial vehicle flight data preprocessing, wherein the data of x, y and z axes transmitted back by a lower computer are shown in table 1:
TABLE 1
According to the statistical principle, the first order value X of the data sample i Obeying a normal distribution, according to the 3 sigma criterion, for X i Data in the (EX-3 sigma, EX+3 sigma) interval is considered to be normal data to be reserved, and X is outside the (EX-3 sigma, EX+3 sigma) interval i Considered as abnormal data, which is to be rejected, i.e. X outside the rejection interval i Anomaly data and corresponding x i+1 Thereby realizing the elimination of abnormal data.
Data n samples { x } are returned for a group of unmanned aerial vehicles 1 ，x 2 ，...，x n After uploading the data, calculating a first order value X i The method comprises the following steps:
X i ＝x i+1 -x i (i＝1，2，...，n-1)
X i describesThe change condition of unmanned plane corresponding data in a period of time is calculated according to statistical knowledge, and the change condition is calculated in a certain interval by calculating X i Is used for data elimination.
Unmanned aerial vehicle flight sample x i Is a level difference value data X of (1) i The mathematical expectation of (a) is:
first order difference value data X i The standard deviation sigma of (2) is:
according to the above formula, the first order difference values X of the data X, y and z shown in Table 1 are calculated i 、Y i 、Z i The method comprises the following steps:
X i 、Y i 、Z i describes the change condition of corresponding data of the unmanned aerial vehicle in a period of time by statistics X i 、Y i 、Z i Is used for data elimination.
First-order difference value data X of unmanned aerial vehicle flight samples X, y and z i 、Y i 、Z i The mathematical expectation of (a) is:
first order difference value data X i The standard deviation sigma of (2) is:
the mathematical expectations and standard deviations for the x-axis, y-axis and z-axis, respectively, are substituted above: in the x-axis data, reserving data in a section (-0.0384,0.1378), and eliminating data outside the section; in the y-axis data, reserving data in a section (-0.0811,0.2631), and eliminating data outside the section; in the z-axis data, the data in the interval (-0.1221,0.1228) is reserved, and the data outside the interval is removed.
After the abnormal values are removed, the data are arranged and combined, interpolation smoothing processing is carried out on the missing data by using a Lagrange interpolation method, and finally the repaired data are output.
For a set of data with outliers removed, its arrangement is expressed as (x 0 ，y 0 )，(x 1 ，y 1 )，…，(x k ，y l ) Wherein x is j (j=0, 1,2, …, k-1, k) represents the position of the argument, y j (j=0, 1,2, …, k-,1, k) represents the value of the function at this position, all x j All are different, and the polynomial of the Lagrange interpolation can be obtained according to the Lagrange interpolation formula:
wherein each l j (x) For the Lagrangian base polynomial, the expression is:
lagrangian base polynomial l j (x) Is characterized in that x is j Take up a value of 1 for x of i.noteq.j i The upper value is 0.
And estimating the data with the outlier removed by using the fitted Lagrangian polynomial, and inputting the data position of the missing value into the polynomial to predict the estimation of the function at the position, thereby realizing the data interpolation estimation of the missing data position.
With the above method, the table 1 data is preprocessed, and data visualization diagrams before and after preprocessing are drawn, as shown in fig. 3 (a) and 3 (b).
Step 3: unmanned aerial vehicle motion feature construction
A dimension-reducing algorithm can reduce multiple indexes into a plurality of main components, the main components are formed by linearly combining original variables, and the variables are independent, so that most useful information in the original data can be embodied. In the invention, a principal component analysis method is adopted to perform data dimension reduction on the data collected by the unmanned aerial vehicle.
The unmanned aerial vehicle flight data are composed of 10 indexes, namely x, y, z-axis speed, x, y, z-axis acceleration, x, y, z-axis Euler angle and unmanned aerial vehicle corner, and an 80 x 10 matrix x is formed by 80 groups of time series samples and is expressed as:
calculating the mean value of matrix x by columnThe method comprises the following steps:
calculate standard deviation S j The method comprises the following steps:
data normalization is performed as follows:
normalized data X ij And (5) conforming to normal distribution of the standard to obtain a standardized matrix X of the original sample.
Calculating a covariance matrix R, and solving eigenvalues and eigenvectors of the covariance matrix RRespectively denoted as lambda i (i=1, 2,., p) and a i (i＝1，2，...，p)。
The contribution rate c of the ith component to the total components i The definition is as follows:
the cumulative contribution rate is defined as:
the contribution rate c is firstly calculated i Sequencing from big to small, and sequentially accumulating the contribution rates to calculate an accumulated contribution rate C i Will accumulate the contribution rate C i The first more than 90% of the components are included as the main components.
According to the invention, the collected unmanned aerial vehicle flight data is analyzed, and the accumulated contribution rate is calculated. Through principal component analysis, the contribution rate C is accumulated i Since the number of components covered when the first time exceeds 90% is 7, the components are selected and retained as main components, and the contribution rate and the cumulative contribution rate of each main component are as shown in fig. 4.
Step 4: unmanned aerial vehicle flight mode identification;
the unmanned aerial vehicle flight status classification is shown in fig. 4. For a given one of the two classified data setsWherein y is i E { +1, -1}, there is one hyperplane:
ω T x+b＝0
the two types of samples are separated, then there is y for each sample i (ω T x+b) > 0, the distance of the sample to the segmentation hyperplane is:
where ω is a hyperplane parameter.
Defining an interval gamma as the shortest distance of all samples in the whole dataset to the segmented hyperplane:
γ＝min γ i (i＝1，2，...，N)
finding out a most suitable hyperplane to classify data, when the hyperplane meets the interval gamma to be maximum, dividing two data sets of hyperplane segmentation is the most stable and has strong robustness, namely solving an optimization equation:
wherein the method comprises the steps ofSatisfies the hyperplane parameter ω of ω x γ=1, the optimization equation is simplified as:
machine learning was performed by the improved SVM model using 40 of the 60 sets of data in table 1 as a training set for machine learning. Using 20 sets of 60 sets of data as a test set for machine learning, the loss function uses a 0-1 loss function, and the loss function between the true distribution y and the predicted distribution f (x; θ) of the tag is:
defining the number of errors of the classifier as:
wherein m is the number of samples.
The exact number of classifiers is the total number of classifications-the number of errors. The accurate number of climbing exercise training is 40, and the accurate number of testing is 20 after calculation; the accurate number of the training of the flat flight exercise is 40, and the accurate number of the test is 20; the accurate number of convolutions is 38, and the accurate number of tests is 17; the accurate number of turns is 37, and the accurate number of tests is 16; the drop accuracy number is 40, and the test accuracy number is 40.
Definition accuracy is:
in the above formula, m is the number of samples, and n is the number of classification errors.
The classification results are shown in table 2:
table 2 classification results table for five states
As shown in table 2, the SVM classifier has a good classification effect in climb, fly-flat and descent recognition, and has a reduced classification effect in a spiral motion state. The classification result shows that the support vector machine model based on the binary tree improvement method is feasible in solving the multi-flight mode classification problem.
Step 5: unmanned aerial vehicle flight path prediction
The BP network topology diagram of the unmanned aerial vehicle track prediction model established by the invention is shown in figure 4. The unmanned aerial vehicle track prediction neural network selects 7 input variables and 4 output variables. The 7 input variables are the positions (longitudinal displacement, transverse displacement and vertical displacement) in the directions of x, y and z coordinate axes, angular motion parameters (pitch angle, roll angle and yaw angle) and heading; the four output parameters are displacement changes in the x, y and z axis directions and time, and the 4 variables are used for describing a 4D track of unmanned aerial vehicle flight.
The empirical formula and the error square sum confirm the hidden layer node calculation formula as
In the invention, the number of the hidden layer neurons is 1; the number of neurons of the hidden layer is 8, and the requirement that the number of the hidden layer is 4-14 is met.
In neural network training, it is necessary to divide the sample data set into a training set and a test set. The training set is used for training the neural network, and the test set is used for evaluating the advantages and disadvantages of the trained neural network. The neural network sample set respectively collects 100 groups of samples according to five movement modes, 80 groups of samples are divided by a training set according to the neural network training principle, and 20 groups of samples are divided by a test set to carry out neural network training. To obtain a reliable model, the learning rate lr was chosen to be 0.001 by multiple trials and experience. The prediction model parameters and the initialization configuration are shown in table 3.
Table 3 neural network parameter settings
Respectively analyzing five flight states of the unmanned aerial vehicle, wherein a predicted track in a flat flight state is shown in fig. 6; the predicted track in the turning state is shown in fig. 7; the predicted track in the climbing state is shown in fig. 8; the predicted track in the descent state is shown in fig. 9; the predicted track in the hover state is shown in fig. 10.
Defining the distance d between the predicted track point and the actual track point as follows:
in the above formula, x represents the actual track x-axis distance of the unmanned aerial vehicle,representing the predicted track x-axis distance of the unmanned aerial vehicle, and y representing the actual track y-axis distance of the unmanned aerial vehicle,/->Representing unmanned aerial vehicle pre-launchMeasuring the distance of the Y axis of the flight path, wherein z represents the distance of the Z axis of the actual flight path of the unmanned aerial vehicle, +.>Representing the predicted track z-axis distance of the unmanned aerial vehicle.
The error μ defining the predicted track is:
the errors of the tracks in the five motion states are calculated respectively, and a track error diagram is drawn as shown in fig. 11. As shown in fig. 11, the error of the unmanned aerial vehicle prediction in the descending state is minimum, the error distance is 0.13m, the error of the unmanned aerial vehicle prediction in the spiral motion state is maximum, and the error distance is 0.41m. The unmanned aerial vehicle track prediction error is within 0.5m, and the test result shows that the unmanned aerial vehicle track prediction model based on the flight pattern recognition can realize the unmanned aerial vehicle track prediction.
Claims (7)
1. The unmanned aerial vehicle track prediction method by using the flight mode identification is characterized by comprising the following steps of:
step 1: collecting and analyzing airborne data of the unmanned aerial vehicle;
step 2: preprocessing unmanned aerial vehicle track data;
because noise exists in the unmanned aerial vehicle track data, data preprocessing is needed before unmanned aerial vehicle data analysis, abnormal data is removed in the data preprocessing, interpolation processing is carried out on the missing data, and data is output after the data preprocessing flow is completed;
step 3: constructing the motion characteristics of the unmanned aerial vehicle;
the principal component analysis method is used as a dimension reduction algorithm, reduces a plurality of indexes into a plurality of principal components, is formed by linearly combining original variables, does not have any relation, and reflects most of useful information in the original data;
step 4: unmanned aerial vehicle flight mode identification;
by finding hyperplane splitting dataImplementing data classification for a given one of the classified data setsWherein y is i E { +1, -1}, if the two samples are linearly separable, there is one hyperplane: omega T x+b＝0；
The two types of samples can be separated, then there is y for each sample i (ω T x+b) > 0, the distance of the sample to the segmentation hyperplane is:
wherein ω is a hyperplane parameter;
defining γ as the shortest distance from all samples in the dataset to the segmentation hyperplane, then:
searching a most suitable hyperplane to classify data, and when the hyperplane meets the maximum interval gamma of all samples, dividing two data sets of the hyperplane segmentation is most stable and has strong robustness, namely solving an optimization equation:
wherein the method comprises the steps ofSatisfies the terms of ω x γ=1, i=1, 2, n., optimization equation reduction is:
the support vector machine model can realize classification recognition; the flight states of the unmanned aerial vehicle comprise five flight states of flat flight, turning, climbing, descending and spiraling, and the flight data of the unmanned aerial vehicle covering the five flight states are subjected to four-time two-classification, namely, the flight states of the unmanned aerial vehicle are subjected to multi-classification identification based on a support vector machine model improved by a binary tree, so that the five flight states of the unmanned aerial vehicle, namely, the climbing, flat flight, turning, spiraling and descending of the unmanned aerial vehicle can be accurately identified, and the unmanned aerial vehicle flight mode identification is realized;
step 5: predicting the flight path of the unmanned aerial vehicle;
establishing a BP neural network topological graph of an unmanned aerial vehicle track prediction model, wherein the unmanned aerial vehicle track prediction neural network selects 7 input layer variables and 4 output layer variables, and the 7 input layer variables are position longitudinal displacement, transverse displacement and vertical displacement in x, y and z coordinate axis directions, and also comprise an angular motion parameter pitch angle, a roll angle, a yaw angle and a course; the 4 output layer variables are displacement changes in the x, y and z axis directions and time, and the four output variables are adopted to describe a 4D track of the unmanned aerial vehicle;
and calculating the flight information of the unmanned aerial vehicle by the trained learning network to obtain a prediction result of the unmanned aerial vehicle flight path, thereby realizing unmanned aerial vehicle flight path prediction of flight pattern recognition.
2. The method for unmanned aerial vehicle track prediction by flight pattern recognition according to claim 1, wherein:
in the step 2, the specific steps are as follows:
(1) removing abnormal data;
data n samples { x } are returned for a group of unmanned aerial vehicles 1 ，x 2 ，...，x n After uploading the returned data, calculating a first order value X i The method comprises the following steps: x is X i ＝x i+1 -x i i＝1，2，...，n-1；
X i Describes the change condition of corresponding data of the unmanned aerial vehicle in a period of time by statistics X i The statistical characteristics of the data are removed, and the removing steps are as follows:
unmanned aerial vehicle flightSample x i Is a level difference value data X of (1) i The mathematical expectation of (a) is:
first order difference value data X i The standard deviation sigma of (2) is:
according to the statistical principle, the first order value X of the data sample i Obeying a normal distribution, according to the 3 sigma criterion, for X i Data lying within the (EX-3 sigma, EX+3 sigma) interval is considered normal data and is retained for X lying outside the (EX-3 sigma, EX+3 sigma) interval i Considered as abnormal data and culled, i.e. X outside the culling interval i Anomaly data and corresponding x i+1 Thereby realizing the elimination of abnormal data;
(2) missing data interpolation
The method comprises the steps of arranging and combining data after abnormal data are removed, carrying out interpolation smoothing processing on missing data by using a Lagrange interpolation method, and finally outputting repaired data, wherein the method comprises the following specific steps of:
the data arrangement after the outlier removal is expressed as (x) 0 ，y 0 )，(x 1 ，y 1 )，...，(x j ，y j )，...，(x k ，y k ) Wherein x is j Representing the position of the argument, y j Representing the value of the function at this position, j=0, 1,2 …, k-1, k, and all x j All are different, and according to the Lagrange interpolation formula, the Lagrange interpolation polynomial is obtained as follows:
wherein each l j (x) Is BraggThe Langerhans' basic polynomial has the expression:
lagrangian base polynomial l j (x) Is characterized in that x is j Take up a value of 1 for x of i.noteq.j i The upper value is 0;
and using the fitted Lagrangian polynomial as the estimation of the data with the outlier removed, and inserting the data position of the missing value into the Lagrangian polynomial to predict the value of the function at the position, thereby realizing the data interpolation estimation of the missing data position.
3. The method for unmanned aerial vehicle track prediction by flight pattern recognition according to claim 1, wherein:
in the step 3, performing data dimension reduction on the data acquired by the unmanned aerial vehicle by adopting a principal component analysis method;
p indexes and n groups of samples of the preprocessed unmanned aerial vehicle flight data form an n multiplied by p matrix x as follows:
calculating the mean value of matrix x by columnThe method comprises the following steps:
calculated standard deviation S j The method comprises the following steps:
data is standardized, and standardized data are:
the normalized matrix for the original sample is obtained as:
the covariance matrix is calculated as:
wherein r is ij The method comprises the following steps:
solving the eigenvalue and eigenvector of covariance matrix R as lambda respectively i And a i I=1, 2, p. p is the index number of the sampling data, the value range of k is [1, p]；
The contribution rate c of the ith component to the total components i The definition is as follows:
cumulative contribution rate C i The definition is as follows:
first, the contribution rate C i Sequencing from big to small, and sequentially accumulating the contribution rates, namely calculating the accumulated contribution rateThe cumulative contribution ratio i=1, 2..p is the main component, which is the component covered when p exceeds 90% for the first time.
4. The method for unmanned aerial vehicle track prediction by flight pattern recognition according to claim 1, wherein:
in the step 4, the propagation process of the BP neural network algorithm comprises two aspects, namely forward propagation of the input layer information and reverse propagation of the output layer error in sequence;
(1) forward propagation of input layer information
The BP neural network consists of an input layer, an implicit layer and an output layer, wherein the input layer is provided with 7 neurons, the output layer is provided with 4 neurons, and the implicit layer takes 4-14 neurons; the forward propagation of the information of the input layer is realized by transmitting the data weighting operation of the input layer to the hidden layer and then transmitting the data weighting operation of the hidden layer to the output layer, thereby realizing the forward propagation of the information of the neural network;
let the connection weight between the input layer (I) th neuron and the hidden layer (h) th neuron be v Ih The connection weight between the h neuron of the hidden layer and the J neuron of the output layer is omega hJ Then the input received by the h neuron of the hidden layer isThe input received by the J-th neuron of the output layer is +.>
Let the transfer function from the input layer to the hidden layer be f 1 Implicit layer to output layer transfer function of f 2 The method comprises the steps of carrying out a first treatment on the surface of the Thus, the output of the hidden layer node is f 1 α h The output result of the output layer node is f 2 β J ；
(2) Counter propagation of output layer errors
Neural network parameter learning using a random gradient descent method for a given i-th set of samples (x i ,y i ) Inputting the sample into a neural network model, and obtaining network output as follows:
(a) Loss function
The loss function is used to quantify the difference between model predictions and real labels, defining a square type loss function as:
the total error generated for the p sets of neural network samples is defined as:
(b) Empirical risk minimization criteria
The smaller the model predicted loss, the stronger the model's predictive ability, in order to obtain a smaller expected error, the Empirical Risk (Empirical Risk) is calculated, i.e. the average loss of the training set is:
(c) Random gradient descent method
And optimizing an experience risk function by using a random gradient descent method, and continuously iterating the algorithm by training the random gradient descent method until the set precision is met, wherein the random gradient descent is converged to a local optimal solution after the iteration is ended.
5. The method for unmanned aerial vehicle track prediction by flight pattern recognition according to claim 4, wherein:
the training steps of the random gradient descent method are as follows:
step 1: input training setVerification set V and learning rate α;
step 2: carrying out random initialization of a sample set;
step 3: calculation ofObtaining a gradient;
step 4: calculation ofUpdating the parameter θ with the result as the initial value of θ, i.e., +.>
Step 5: repeatedly performing step 3 and step 4 continuously until the model f (x i The method comprises the steps of carrying out a first treatment on the surface of the θ) when the error rate of the verification set V is the lowest, the iteration ends, and the loop is jumped out;
step 6: and outputting theta.
6. An unmanned aerial vehicle on-board data acquisition system according to the predictive method of any one of claims 1 to 5, wherein:
the unmanned aerial vehicle airborne data acquisition system comprises a lower computer, a wireless transmission module and an upper computer, wherein the lower computer comprises a power supply module, a power module, a GPS positioning module, a flying height measuring module, an attitude detecting module and an STM32F427 singlechip main control module, the wireless transmission module is built by adopting a 3DR wireless data transmission module, and a wireless communication channel is provided for wireless communication between the upper computer and the lower computer; the upper computer builds a monitoring system by LabVIEW, and realizes the functions of monitoring, controlling and recording airborne data in the flight process of the unmanned aerial vehicle.
7. The unmanned aerial vehicle on-board data acquisition system of claim 6, wherein:
in the unmanned aerial vehicle airborne data acquisition system, a power supply module provides power for a STM32F427 singlechip main control module and is matched with a power module to provide power for normal flight of the unmanned aerial vehicle, and a GPS positioning module measures the flight longitude and latitude of the unmanned aerial vehicle; the flying height measuring module is used for measuring the flying altitude of the unmanned aerial vehicle; the gesture detection module is used for measuring the flight acceleration and the angular speed of the unmanned aerial vehicle; the flight longitude and latitude, the flight altitude, the flight acceleration and the angular velocity form real-time flight information of the unmanned aerial vehicle, the real-time flight information is transmitted to an STM32F427 singlechip main control module, the STM32F427 singlechip main control module realizes wireless acquisition of unmanned aerial vehicle data through a wireless transmission module, the data is transmitted to a LabVIEW upper computer through a serial port, and the upper computer displays the flight altitude, the flight speed, the flight distance, the rotating speed of a propeller, the electric quantity of a battery and the flight attitude of the unmanned aerial vehicle in real time and records and stores real-time positions, flight tracks and historical data; and finally, inputting unmanned aerial vehicle-mounted information recorded and stored by the acquisition system into MATLAB software for subsequent data analysis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311576511.3A CN117633640A (en) | 2023-11-23 | 2023-11-23 | Unmanned aerial vehicle track prediction method based on flight mode identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311576511.3A CN117633640A (en) | 2023-11-23 | 2023-11-23 | Unmanned aerial vehicle track prediction method based on flight mode identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117633640A true CN117633640A (en) | 2024-03-01 |
Family
ID=90033138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311576511.3A Pending CN117633640A (en) | 2023-11-23 | 2023-11-23 | Unmanned aerial vehicle track prediction method based on flight mode identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117633640A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104950906A (en) * | 2015-06-15 | 2015-09-30 | 中国人民解放军国防科学技术大学 | Unmanned aerial vehicle remote measuring and control system and method based on mobile communication network |
| CN107065902A (en) * | 2017-01-18 | 2017-08-18 | 中南大学 | UAV Attitude fuzzy adaptive predictive control method and system based on nonlinear model |
| CN110610512A (en) * | 2019-09-09 | 2019-12-24 | 西安交通大学 | Unmanned aerial vehicle target tracking method based on BP neural network fusion Kalman filtering algorithm |
| CA3067573A1 (en) * | 2019-01-14 | 2020-07-14 | Harbin Engineering University | Target tracking systems and methods for uuv |
| US20210043098A1 (en) * | 2019-08-06 | 2021-02-11 | University Of Electronic Science And Technology Of China | Methods and systems of active discovery and collaborative collision avoidance of aircrafts |
-
2023
- 2023-11-23 CN CN202311576511.3A patent/CN117633640A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104950906A (en) * | 2015-06-15 | 2015-09-30 | 中国人民解放军国防科学技术大学 | Unmanned aerial vehicle remote measuring and control system and method based on mobile communication network |
| CN107065902A (en) * | 2017-01-18 | 2017-08-18 | 中南大学 | UAV Attitude fuzzy adaptive predictive control method and system based on nonlinear model |
| CA3067573A1 (en) * | 2019-01-14 | 2020-07-14 | Harbin Engineering University | Target tracking systems and methods for uuv |
| US20210043098A1 (en) * | 2019-08-06 | 2021-02-11 | University Of Electronic Science And Technology Of China | Methods and systems of active discovery and collaborative collision avoidance of aircrafts |
| CN110610512A (en) * | 2019-09-09 | 2019-12-24 | 西安交通大学 | Unmanned aerial vehicle target tracking method based on BP neural network fusion Kalman filtering algorithm |
Non-Patent Citations (1)
| Title |
|---|
| JIAN-DONG ZHANG 等: "UAV Trajectory Prediction Based on Flight State Recognition", IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS, 10 August 2023 (2023-08-10), pages 1 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tisdale et al. | Autonomous UAV path planning and estimation | |
| CN114048889B (en) | Aircraft trajectory prediction method based on long-term and short-term memory network | |
| CN109034376B (en) | L STM-based unmanned aerial vehicle flight state prediction method and system | |
| Nuske et al. | Autonomous exploration and motion planning for an unmanned aerial vehicle navigating rivers | |
| CN110866887A (en) | Target situation fusion sensing method and system based on multiple sensors | |
| CN104765968B (en) | A kind of UAS health status assessment device | |
| Mammarella et al. | Machine vision/GPS integration using EKF for the UAV aerial refueling problem | |
| CN113869379A (en) | Data-driven aircraft energy anomaly identification method | |
| CN111460676A (en) | Method for evaluating health performance of multi-rotor aircraft under atmospheric turbulence disturbance | |
| Hentati et al. | Mobile target tracking mechanisms using unmanned aerial vehicle: Investigations and future directions | |
| Minghui et al. | Deep learning enabled localization for UAV autolanding | |
| Zhang et al. | UAV Trajectory Prediction Based on Flight State Recognition | |
| Silva et al. | Landing area recognition by image applied to an autonomous control landing of VTOL aircraft | |
| CN117311393B (en) | Unmanned aerial vehicle autonomous flight path planning method and system | |
| Kant et al. | Long short-term memory auto-encoder-based position prediction model for fixed-wing UAV during communication failure | |
| Lu et al. | Perception and avoidance of multiple small fast moving objects for quadrotors with only low-cost RGBD camera | |
| CN116954264B (en) | Distributed high subsonic unmanned aerial vehicle cluster control system and method thereof | |
| Pytka et al. | Measurement of aircraft ground roll distance during takeoff and landing on a grass runway | |
| Shi et al. | Design of UAV Flight State Recognition System for Multi-sensor Data Fusion | |
| Kong et al. | A test and evaluation framework for unmanned surface vehicle | |
| CN117633640A (en) | Unmanned aerial vehicle track prediction method based on flight mode identification | |
| Campbell et al. | Loss of control detection for commercial transport aircraft using conditional variational autoencoders | |
| Qi et al. | Detection and tracking of a moving target for UAV based on machine vision | |
| Guan et al. | A new integrated navigation system for the indoor unmanned aerial vehicles (UAVs) based on the neural network predictive compensation | |
| Shi et al. | Design of a UAV Trajectory Prediction System Based on Multi-Flight Modes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |